On 11/17/2009 01:59 AM, Mahajan Vivek-B08308 wrote:
From: Grant Grundler [mailto:grund...@google.com]
Sent: Monday, November 16, 2009 11:08 PM
+static int sata_sil24_msi; /* Disable MSI */
+module_param_named(msi, sata_sil24_msi, bool, S_IRUGO);
+MODULE_PARM_DESC(msi, "Enable MSI (Default: false)");
Vivek,
Do we even still need the parameter? I'm thinking either MSI
works with a chipset or it doesn't. The kernel has globals to
"know" which state is true.
Sometimes even in a platform, some PCIe endpoints do very
well with MSI while others may have to resort to legacy ints.
Should we let the endpoints make the final call.
If the parameter is needed, when this driver is compiled into
the kernel, how is "msi" parameter specified?
I think the parameter needs to be documented and fit in with
other "msi" parameters.
See "nomsi" in Documentation/kernel-parameters.txt.
In this case "msi" is supposed to be passed via insmod and
not via kernel cmdline. If the driver is built-in the kernel,
then force sata_sil24_msi = 1 in the driver to enable it.
First, the original patch was just fine, and it was applied. You should
have received email confirmation of this already.
Second, all module options are available on the kernel command line,
when a module is built into the kernel. You supply a module name prefix
to each module option, on the kernel command line.
Jeff
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
